| Abstract
| - We present a new method for constructing discriminating substructures by reassemblingcommon medicinal chemistry building blocks. The algorithm can be parametrized to meetdiffering objectives: (1) to build features that discriminate for biological activity in a localstructural neighborhood, (2) to build scaffolds for R-group analysis, (3) to construct clustersignatures that discriminate for membership in the cluster and provide a graphical representation for its members, and (4) to identify substructures that characterize major classes in aheterogeneous compound set. We illustrated the results of the algorithm on a literature datasetis of 118 compounds with in vitro inhibition data against recombinant human protein tyrosinephosphatase 1B (PTP-1B).
|
